Large-Scale Network-Oriented Protection Algorithm Through Efficient Tree Granularity Brownian-Like Motion

Abstract

In extensive and decentralized systems with numerous end-to-end communications, inadvertent disconnection can lead to the complete severance of the upper-level applications. Reliable communication can be ensured by designating links as trunks. These trunks maintain connectivity even when a large number of nontrunk links are simultaneously severed. Due to the computational complexity of cut enumeration, prior algorithms are restricted to topologies with a limited number of vertices. To address this challenge, in order to efficiently search in the potential tree space containing all cuts, we search at the granularity of the tree in a tree space that is homomorphic to the cut space and effectively discover a large number of target cuts by combining with Brownian-like motion. The found quasi-trunk links are then filtered to select a minimal-cost subset of trunk links, thereby guaranteeing the desired sustained connectivity. Experimental results demonstrate that the algorithm achieves a three-order of magnitude acceleration compared to the optimal method in small topologies, with limited extra cost. In large topologies, our trunk assignment method exhibits resilience to at least 99.9% stochastic link failures within a reasonable execution time.